Consider the delivery time data discussed in Example 11.3. Find the PRESS statistic for the model developed from the estimation data. How well is the model likely to perform as a predictor? Compare this with the observed performance in prediction.

Data From Example 11.3

Transcribed Image Text:

i All 40 observations for the delivery time data in Examples 3.1 and 11.2 are shown in Table 11.3. We will assume that these 40 points were collected at one time and use the data set to illustrate data splitting with the DUPLEX algorithm. Since the model will have two regressors, an equal split of the data will give 17 error degrees of freedom for the estimation data. This is adequate, so DUPLEX can be used to generate the estimation and prediction data sets. An x-X2 plot is shown in Figure 11.1. Examination of the data reveals that there are two pairs of points that are near Observation, Cases, Distance, x X2 Delivery Time, y Estimation (E) or Prediction (P) Data Set 11 12 733467278569 1234567890 560 16.68 P 220 11.50 P 340 12.03 P 80 14.88 150 13.75 330 18.11 110 8.00 210 17.83 30 1460 79.24 605 21.50 16 688 40.33 10 215 21.00 13 4 255 13.50 14 6 462 19.75 15 448 24.00 16 17 18 19 20 21 22 23 24 25 26 27 28 30 1067370606427556 776 29.00 200 15.35 132 19.00 36 9.50 770 35.10 140 17.90 810 52.32 9 450 18.75 8 635 19.83 150 10.75 22 905 51.00 520 16.80 290 26.16 500 19.90 1000 24.00 31 6 225 18.55 32 10 775 31.93 33 212 16.95 34 1 144 7.00 3 126 14.00 3. 126 14.00 12 10 200 655 37.03 420 18.62 7 150 16.10 82 360 24.38 40 32 1530 64.75 BBBBBEEPPEPEPPEREEEEEEPEREEEPPPPPPPPPP